面向无人机红外影像拼接的特征提取算法对比研究

Feature Extraction Algorithm for UAV Infrared Image Mosaic

由于无人机获取的红外影像具有低对比度、大几何畸变、大影像间倾角以及存在噪声影响等特点,使得无人机红外影像拼接颇为困难。这里首先采用SIFT,SURF及Cen Sur E 3种尺度不变特征检测算子分别对无人机红外影像进行特征匹配实验,验证了Cen Sur E算子在红外影像特征提取中的优越性。最后构建了基于Cen Sur E算子的拼接流程,用同一航带5张无人机红外影像进行拼接实验,得到了满意的拼接效果。

Since the infrared images acquired by UAV have the characteristics of low contrast, large geometric distortion, big inclination and noise, the UAV infrared image mosaic is quite difficult. Firstly, the feature matching experiments of UAV infrared image were done by using the three scale-invariant feature detection operators of SIFT, SURF and Cen Sur E respectively. The experiments verified that the Cen Sur E operator had advantages in the infrared image feature extraction. Finally, the image mosaic process was designed based on the Cen Sur E operator.The image mosaic experiments were done by using the five UAV infrared images in a flight strip and the satisfactory image mosaic results were achieved.